Intel’s (NASDAQ:INTC) software division has gone from a running joke to inexplicably bad, it is almost pitiable now. With the long overdue and badly messaged knifing of Meego, and support for the Linux Foundation’s Tizen project, you can only shake your head.
Intel’s software attempts, with the notable exception of the tools division, has been a complete joke for years. Their drivers are either dull but mostly working or functional in name only, but both are antagonistic to their intended user bases. There may be far worse out there, Creative comes to mind, but there are few that shout as loudly, and none that spend what Intel does. Intel likely spent millions of dollars for each copy of their OSes sold to date.
Today, Intel dumped their latest in-house OS, Meego, and jumped in behind the Linux Foundation’s Tizen project. So much for Tizen going anywhere, up until this point, it looked promising. Why are we so down on Intel and software? Mainly because of their graphics ineptitude, and how key that is to the future of computing. Lets look at their history in this space, it is rather sobering.
The current Intel graphics line is derived from the maligned i965G line, known internally as ‘Gen’ graphics. The i965G never actually worked. No, really, the features Intel promised at launch, the main selling points for the chipset, hardware texturing and lighting for DX9 never worked. The promised DX10 support never came. (Note: It did come out for one obscure mobile variant, but functional and tolerable weren’t adjectives often coupled with the result.)
Intel’s follow up, the G35, got DX10 drivers, or DX10ish drivers, about a month before it was replaced by the G45. The G45 was, well, slow, buggy, and about as worthless as you could inflict on a customer and not be sued for intentional infliction of emotional distress. The replacement was the GPU in the Westmere/Core iSomethingmeaningless with a lower confusing number than is currently out. It’s performance was…. already forgotten for good reason.
That brings us to Sandy Bridge, and the supposedly ‘good’ graphics that this part has. It sucks. 9 months later, the drivers barely work unless you stick to a modern app that Intel knows big name sites benchmark with. Anything out of line or old….. good luck with basic functionality.
The Linux side is even worse, 9 months on, you simply can’t get any distro with graphics support for Sandy Bridge above barely functional. To make matters worse, Intel’s abjectly stupid policies on the matter are only matched by their ignorant messaging. There is no way to download a graphics driver from Intel for their GPUs. Really. If you go to the page, you get a list of devices that Intel doesn’t actually provide drivers for. Brilliant!
If you follow the links to the page with the graphics work Intel is doing for Linux, here, it looks promising. Until you try and actually get a driver. Good luck getting it to work unless you are intimately familiar with pulling source from a Git tree, compiling it, and manually solving all the listed dependencies. There is no downloadable package, period.
If you bring this up to Intel, which the author has multiple times, they come back with an argumentative response about how they are making open drivers, and if they actually released the drivers in a usable form, they would somehow have to close them off. Other than the fact that the majority of the features are not slated to be released on Linux because Intel won’t actually open them up, this is a completely bogus argument. And Intel knows it. Ironically, this problem doesn’t seem to happen at ATI or Nvidia.
Worse yet, they know that the majority of ‘journalists’ out there don’t understand it, so they will repeat what they are told as ‘fact’. The conclusion I have is that Intel does not actually know how to compile and package their own drivers, every other explanation only makes them look simply evil.
On the Atom side, things are far far worse. Intel started out with a bang by screwing over everyone who bought a non-MS running first generation Atom machine. The graphics drivers provided by Intel don’t work well, are not documented. Worse yet, they are done in such a way that they preclude changes to the system they came with. There have been no updates since, no releases of documentation, and no help for those that want to do something as radical as patching their machines.
When the community objected to this crappy behavior, Intel first blamed them, then tried to deflect the blame, and finally pointed out to the author directly that they just don’t care. Remember this part, it is community building in action.
The current situation was looking like a minor step forward until the release of the current Cedarview 32nm Atoms. Those long delayed DX10.1 parts have lost pretty much all of the design wins touted at this Spring’s IDF in China, and are only now trickling out six months late. Cedarview is too hot and, wait for it, doesn’t have working graphics drivers. Yup, it is the finest DX10.1 part to ever blow out its TDP, except it is not DX10.1 anymore. Yes, Cedarview is now the finest DX9 part to ever blow out its TDP.
Intel is so proud of this new part that it didn’t even issue a press release. God knows the state of the Linux drivers, but it is probably OK because Intel seems to have lost 100% of it’s design wins. If you can’t actually buy one, do working drivers matter? ARM is beating Intel like a drum in this space, and they have working drivers and community support.
Then we come to Larrabee. That chip is so tainted that the working version, C0, has been renamed Knights Corner. We here at SemiAccurate HQ had high hopes for this part, but we will refrain from beating this dead horse at the moment, it is simply too easy and smelly for that matter.
Intel simply can’t make working graphics drivers. Intel can’t stop publicly blaming everyone else for their software ineptitude. Intel can’t actually release any software they promise in the space, drivers, OS, or anything but tools. Intel can’t stop attacking anyone who points out problems with their software. Intel won’t attempt to fix driver problems if there is a new part coming in the near future. Intel will never retroactively fix problems no matter how glaring, ask the Atom owners, i965G owners, and likely Cedarview purchasers. Their software side is broken, politically tainted, and unable to actually function. Intel can’t build a community around their software products, and is moving in the wrong direction.
With this background, you can see why we fear for Tizen. It may be a great thing, but if Intel jumps in behind it, their politics, games, and lack of understanding can only cause problems. Their track record is a 5+ year road map on how not to do things, and sadly, the company is increasingly moving in the wrong direction. The community would probably do better without their participation, just write a large check and send over the technical docs as early as possible. Then don’t call us, we’ll call you. Intel needs to actually come up with a corporate strategy to address the deeper underlying issues, but are utterly failing to do so. In the mean time, ARM is both supporting Tizen and eating Intel’s lunch.S|A
Updated: October 3, 2011 10:45pm. Knights Corner was mis-named “Knights Bridge” which has now been corrected. The next generation is Knights Ferry. Some confusion reigned over ferries going under bridges as well as “that’s all water under the bridge” and of course that leads us to make such corrections.
Latest posts by Charlie Demerjian (see all)
- What does Qualcomm’s server SoC look like - Apr 15, 2015
- How does Qualcomm’s SenseID fingerprint scanner work? - Apr 9, 2015
- How fast is Qualcomm’s 64-bit Kryo server core? - Apr 7, 2015
- Amazon is spending lavishly on game development - Apr 6, 2015
- What is the name of Intel’s Cannonlake +1 server platform - Apr 6, 2015